Power Grab

Podcast Transcript

Ed Coyne: Hello and welcome to Sprott Radio. I'm your host, Ed Coyne, Senior Managing Partner at Sprott Asset Management. I'm pleased today to welcome Alan Howard. Alan is a principal analyst in cloud and data center research practice at Omdia. Alan, thank you for joining Sprott Radio.

Alan Howard: I'm pleased to be here. Thanks for having me, Ed.

Ed Coyne: Alan, before we discuss a topic that is near and dear to your heart as well as Sprott's, the growing relationship between tech and energy, let's start by talking a bit about your background, some of the work you're doing today and the company you work with.

Alan Howard: Sure. The company I work for is called Omdia, but we're just a division inside a much larger organization. Informa Group is based in London and has a broad media and events business. Inside Informa Group is a division called Informa Tech. Omdia is the research arm inside of Informa Tech, and Informa Tech runs things like Data Center World and Black Hat. Those names would resonate immediately if you're in the IT industry, and I've been in the IT industry for 40 years.

I've been studying data centers for much of that time. I've been here at Omdia for a little over six years, and in particular, I study people who run, operate and build data centers. What do the big data center operators' portfolios look like? What kind of data centers are they building? Things of that nature.

Ed Coyne: That's evolved substantially over the last couple of decades. From 40 years ago to today, you've seen tremendous growth. That's probably an understatement, right?

Alan Howard: You're right. Yes.

Ed Coyne: Can you talk about what's happening regarding energy and tech and how those two are being infused together today? We really can't think about one without thinking about the other. Can you walk us through that evolution and what you've witnessed over those four decades of being in this space?

Alan Howard: There's probably a dozen different market dynamics that play into the whole thing, but generally speaking, the evolution of a digital economy has occurred over the last couple of decades. That's both on the business side and the consumer side. I think one of the things is that if you're not really in tune with what a data center is, if you think about your phone, everything you do on your phone touches data centers.

When you transact something, make a phone call, or send a text, it all seems like it goes to a cell tower, and then, I don't know. What happens at the cell towers, generally speaking, is that the signal gets converted into a digital signal. It goes across fiber optic cables and traverses several data centers. Whether it's on your computer or your phone, these data centers process and make all those things happen.

You can imagine how it's grown over time, but I'll throw in two other things that I think are specifically pivotal in evolution. One is that cloud services began to take off. Cloud services started roughly a dozen years ago. You can argue that in different ways, but demand drove up quite a bit as they caught on with the business world.

It was probably about 2018, roughly, when the wheels came off in terms of expanding data center capacity to handle this digital economy. Then, when the pandemic hit, another big catalyst was that people found themselves sheltering in place and needed new services. If you're an analyst studying this market, it's been a wild ride. We've updated a number of our forecasts over and over again as we learn more about new announcements with regard to AI hardware and things like that, so it's a pretty wild ride.

Ed Coyne: You talk about shelter in place; COVID has kickstarted many things that will happen organically over time, such as a more flexible work schedule, working from home, and working in different locations. More data and AI are needed to assist. Did it fast-forward all this by a decade or two? Are we caught off guard as far as energy needs and so forth? What happened because of COVID-19 as far as the trajectory of all this? Did it expedite it, or was it already happening?

Alan Howard: In terms of data centers expanding to handle the additional demand created by COVID-19, yes, we're already in a market growth situation where if you look at the biggest builders of data centers, they are so tooled up with people and processes and so on to expand data centers. The wheels are already in motion.

Now, the rise of AI has been yet another catalyst that will drive additional expansion. As importantly, it's also creating a new set of companies looking to ride this wave's coattails and create data center campuses. The big guys, Amazon, Google, Meta, and Microsoft, are hoping to land some of that business, so a number of different things are going on.

Ed Coyne: Those are the big players, as you mentioned. The Microsofts, the Amazons, and so forth of the world here in the U.S. I was surprised to see that they equate to maybe, I guess, a couple hundred, maybe 500 total data centers, but there's over 10,000, if I'm reading that correctly, globally. That must be a major draw on our energy and the resources we need today. Is that number a real number, or is that inflated?

Alan Howard: The 10,000?

Ed Coyne: Yes.

Alan Howard: That’s way low, but let me characterize it. We look at it from a power capacity perspective, so how much power capacity is at data centers now, and how will that grow over time? I want to make a quick point: there are two conversations. There's a conversation about power capacity. How much capacity do I need to meet peak loads when they occur?

I find that many people I speak to are often talking about power consumption. Power consumption is a different subject. It's the amount of power that gets used, what I pay for my utility, but power capacity is what we have to build up to be sure we can meet those peak loads when everybody's shopping on Amazon for Christmas.

Ed Coyne: Between real estate and energy creation and consumption, is that becoming a bigger topic, or is that overblown right now, given where we are in the universe? What's the reality behind that?

Alan Howard: That's a good question because the reality behind it is in the data center industry, there are big builders like Amazon, Google Meta, and Microsoft. I would love to give you the whole history. Still, I'll summarize it by saying that Greenpeace started putting out reports, I think, in 2014, picking on tech industries for the amount of power they were consuming and how much greenhouse gases they emitted.

That catalyst started driving the industry to a very high focus on sustainability. When we speak about what I refer to as hyper scalers, which are Amazon, Google, Meta, and Microsoft, and you look at their sustainability reports, they're very transparent nowadays. Sustainability is at the top of their list. Generally speaking, you can't run a data center without grid power. Almost all the big hyperscale cloud folks have a high focus on it.

Ed Coyne: Everything I've come across seems to be focused on not just the sustainability but also the dependability of these things. Data centers and AI have to be, by definition, a 24/7 business. They can't have brownouts, rolling brownouts and so forth. For some reason, it's okay for our home to have a brownout, but you can't have these things have a brownout.

If my phone stops working for 10 minutes, I panic. I can't imagine what someone in their teens or 20s would do if their phone stopped working for a few minutes. How are these data centers trying to tackle that? Because you mentioned that they all live off the grid. As we know, the grid isn't 100% defendable or predictable, so how are they combating that going forward? What kind of backups do they have?

Alan Howard: It's rare to have a data center that doesn't have backup power, so right now, the primary go-to is diesel generators. What's shocking, if you look at aerial views of data centers around the outside of the building, is all this infrastructure, and it's a combination of power and cooling equipment. Still, the number of generators is just staggering. That's evolving, too.

One of the conversations in the industry is how to get to a generator-less data center. That's going to be a long way off, but I talk about diesel generators. One of the evolutions is using a different fuel called HVO that puts fewer greenhouse gases into the atmosphere.

Ed Coyne: Some readings and conversations I've had show that these smaller modular reactors on the nuclear front are coming up in conversations as a potential solution. From your point of view, is that fact or fiction? Are we going to see a point in time where data centers are going to have their own energy sources, where they're creating and storing and consuming energy and effectively owning their nuclear power plant? Is that a scientific, forward-looking, aspirational thing, or are you seeing conversations start to happen around that? You're on the front line. What are you seeing out there?

Alan Howard: I have ongoing relationships with the data center industry and the nuclear industry. We spend quite a bit of time talking to people at Idaho National Labs and some of the other national labs. I think that in the future, it will become commonplace. I don't think there's a choice. The notion of small modular reactors or microreactors makes a lot of sense. However, there's a lot of debate about how soon. When? We see some very positive activity going on, but that positive activity is, generally speaking, test reactors.

The industry is going to need to evolve in the sense that once we get past this test reactor phase and now it's commercially viable to take an SMR and put it somewhere and create a base load source for whether it's data centers or steel mills or other industrial applications, that's going to be probably ten years. Let's say we have SMRs ready to be built for commercial application in five years. It'll be another number of years before the supply chain, the economics, everything else comes together.

In April, we ran a workshop at Data Center World in Washington, DC. It was an all-morning workshop. We had great attendance. We had 200 to 250 people from the data center community in the room. Primarily, a couple of representatives from Idaho National Labs spoke. We had been working to develop this for a long time because what wasn't happening was different industries getting together. The nuclear industry has been off on its own for a long time. Research and development has been going on for a darn long time.

Essentially, SMRs have been in submarines for 60 or 70 years. What brought all these people together was the realization that it may be five years before it's commercially viable and ten years before it's broadly commercialized. However, the time to start talking to the industries that will be the consumers is now. You have nuclear in the government, you have the private nuclear research and development folks, and then you have the consumers, whether it's data centers or other industries.

To have a viable business model and reasonable economics, the argument is these industries need to be talking to each other now, understanding how things work, understanding I'm operating my business with these kinds of margins based on this kind of power generation. I want to be using green power, but we'll need the business models of these nuclear companies to evolve in a way that makes sense for the consumers.

We do power forecasts. Obviously, we do all kinds of forecasts, and that's what a research company does. In 2023, global data center capacity was about 100 gigawatts. When I first started talking to Idaho National Labs a couple of years ago, they were shocked when I told them that.

"Oh my God, you're kidding. We need to start talking to some of these people." The forecast for going out in 2030 is 230 gigawatts. That's crazy.

My point is that the forecast is there, given everything falls into place just like it should. There's a lot of market dynamics and moving parts at play. What if the supply chain can't handle delivering all that AI hardware? What if the power grids in various countries can't support it? There are a lot of variables in there. You can do a forecast like that. If all the ducks line up, then that's a tremendous amount of power. I forget exactly what it is for the U.S., but it's big. If you think about a large-scale nuclear power plant, you're talking a little over 3 gigawatts in just rough numbers. It takes ten years to build a nuclear power plant. That's not going to solve this problem.

Power is a big issue. One of the things that we read a lot about in the press is that we're running out of power here and there. There are power-constrained markets, but I think one of the things that people need to understand is that it's not that we're running out of power in the U.S.; it's that people keep demanding more power in strained markets.

If you think about these hyperscale guys, because Amazon, Google, Meta, and Microsoft build a tremendous amount of data centers, they're really big, but you know what? They build many of them out in the middle of nowhere, Nebraska, Iowa, Kansas, and so on. When we talk about whether there is going to be enough power for all this AI stuff, there's power out there if you start learning more about the electric grid and transmission and regulation and the fact that there are different areas around the country where if I have excess energy, I can't give it to one of my neighbors because there's no transmission infrastructure or perhaps there's some regulatory issues.

One interesting thing is that if I want to build a data center, I have to have a parcel of land with three characteristics. It has to have power and reasonable access to network fiber, and as a general rule, it will have to have water. That trifecta right there is complicated. One of the things that we've seen, and we've seen it for a long time now, is that these companies are buying parcels of land that make sense for building a data center and land banking them—a ton.

Ed Coyne: What's land banking? I haven't heard the term.

Alan Howard: Land banking is: I'm buying a parcel here because it has all the right characteristics, but I don't need it. Why would I buy it if I don't need it? Well, if that has all the right characteristics, somebody else will probably find it. They're banking all this land, so the big cloud guys that build data centers and the big co-location guys that build data centers acquire this land because the land is a very small percentage of the cost of building a data center. When we build data centers, we talk about it in terms of power. What does it cost to build a data center? It costs 8 to 12 million a megawatt. That's how important power is.

Ed Coyne: I love that quote. Power is life. It's such a true and direct quote.

Alan Howard: I hope to see how this AI evolves before I retire.

Ed Coyne: You and me both. How can someone keep track of you, follow you, and follow the work you guys are doing? What resources would you recommend for those that we've piqued their interest and want to keep track of this evolution?

Alan Howard: Actually, LinkedIn is probably the best option. If you're interested in this stuff, there's a website, just omdia.com, with blogs about many of these subjects.

Ed Coyne: This is great. Alan, I appreciate you taking the time. I wrote a ton of notes, and I learned along the way. I thought I prepared well for this but learned even more, so this was great. Thank you all for listening. Once again, I'm Ed Coyne, your host, and thank you for listening to Sprott Radio.